Logistics distribution and inspection of unmanned vehicles based on improved YOLOv5

With the continuous development and application of unmanned vehicle technology, more and more unmanned vehicles must work in a variety of different bad weather and environmental conditions, which also brings higher requirements to the research of unmanned vehicle perception systems. Especially in foggy days or other severe weather conditions, the perception of complex road states becomes more difficult. Therefore, to improve the perception effect of unmanned vehicles in bad weather, this paper proposes a multi-functional unmanned vehicle visual perception system based on YOLOv5. This paper proposes three aspects to improve the problem of the perception effect of unmanned vehicles in bad weather. First, the model balance between computational efficiency and accuracy is improved by including the Ghost Bottleneck module. Secondly, the CBATM module is used to enhance the target perception ability of the model, especially the detection accuracy in foggy scenarios. Finally, the MSR algorithm is combined to enhance the robustness of the model in foggy scenarios and improve the ability of the model to perceive targets in a complex environment. This YOLOv5-based multi-functional unmanned vehicle visual perception system has a wide application prospect in the application of multifunctional unmanned vehicles integrating distribution and inspection and provides strong support for the realization of intelligent perception and decision-making.